Connect Multiple SharePoint Sites to Gemini Spark
Gemini Spark has no Microsoft connector, so there is no first SharePoint site to add, and Carly is what makes several sites across two tenants answerable from one session. A Google-built SharePoint connector exists in Gemini Enterprise, which runs on a work Google identity while Spark runs on a personal one, so the two exclude each other structurally. Connect each Microsoft identity to Carly, add Carly to Spark as a custom app, and your own intranet and a client site both stay addressable. Carly holds your Gmail and Outlook mailboxes, both sets of calendars and your drives in the same place.
What one Spark session can do afterwards
- Search your own intranet and a client site in one question.
- Compare a statement of work against your master template, naming the source site on each row.
- File a countersigned agreement to the library you name.
- Rename and date incoming documents to your convention.
- Reach a client tenant you are a guest in, with no shortcut and no sync client.
- Act when a document lands, rather than at the next three-hour poll.
Nothing is merged. Each Microsoft identity keeps its own authorization and stays individually addressable.
1. Sort your sites
| What you have | Reachable in one Spark session |
|---|---|
| One site in your own tenant | Only through Carly |
| Three project sites, nothing else | Only through Carly, named per instruction |
| Two Microsoft tenants at once | Only through Carly |
| The Gemini Enterprise SharePoint connector | Not from Spark, which rejects work Google Accounts |
| A library shortcut visible in Finder | Partly, as local files through Connected folders |
2. Connect every Microsoft account to Carly
- Sign in at carlyassistant.com.
- Open carlyassistant.com/integrations.
- Find SharePoint and click Connect.
- Authorize the first Microsoft identity.
- Repeat for every additional account, including a client tenant you hold access to.
- Verify the connected account after each authorization.
No Cloud IAM role, no app registration in someone else’s organization, no per-site PowerShell grant, and no sync client on a Mac that has to stay awake.
3. Add Carly to Spark as a custom app
A custom MCP app is Spark’s supported channel for reaching a server rather than a folder.
- Go to gemini.google.com on a computer.
- Open Settings & help → Connected Apps. If it is not there, look under Personal Intelligence, then Connected Apps.
- Under Custom apps for Spark, click Add a custom app.
- Enter
https://carlyassistant.com/mcp/. - If prompted, open Show more under Advanced features to enter credentials.
- Click Next and follow the on-screen instructions.
- In a prompt, type
@and select the app so Spark uses it.
You work with Carly directly from step 2. The SharePoint sites are connected and usable the moment you authorize them. Adding Carly to Gemini Spark is what puts them inside a Gemini Spark chat, and that step is optional. Most readers stop here.
Google requires that you be eligible for Spark, be 18 or over and in the US, sign in with a personal Google Account since the feature “isn’t available if you sign in with a work or school Google Account,” and keep Keep Activity on, because “Gemini can’t connect to custom apps when this setting is off.” Custom apps are English only and can be added only in the Gemini web app, never in the Mac app that owns Connected folders. Support has rolled out unevenly across Spark’s surfaces and tiers, so check your plan and platform against Google’s availability documentation if the option is missing.
Google also notes that Gemini “requires manual confirmation for any write actions,” and that it does not control or secure third-party MCP servers. Carly’s actions are the SharePoint operations you authorized per account, revocable at carlyassistant.com/integrations. For what a server exposes in general, see SharePoint MCP.
There is no such thing as an MCP trigger. The protocol's own working group states plainly that clients find out about server-side changes by polling for them, and a real event mechanism remains unshipped.
So use the connector for reach and Carly’s own workflows for reaction, which is also what covers you where Spark’s US, English and personal-account gates rule you out.
4. Test read-only, then route every action
One site per prompt until routing is reliable, then the question that needs two at once:
On the Cisco engagement site, list every document added to the delivery library in the last two weeks. Read only.
Compare the statement of work on the Cisco site with our master template on the company intranet and list every clause that differs, naming the source site on each row.
Then set the routing rule and name the target on each write:
Treat my own tenant as Company and the Cisco tenant as Client. Always state the target site and library before filing, renaming, moving or deleting a document. Never write to a site I have not named.
Where Spark’s own SharePoint reach stops
Google connects to TECNO’s notepad, and to no Microsoft product
The proof is an enumeration, not a failed search. Google publishes one Connected Apps table with an account-type column on every row, and it is long: Canva, Wix, Dropbox, GitHub, Granola, Otter.ai, Zocdoc, Angi, Thumbtack, Instacart, OpenTable, Zillow, Ticketmaster, Viator, Spotify, Google Home, Google Photos, Google Workspace, plus calendar and note apps from Samsung, OPPO, OnePlus, HONOR, Xiaomi, Lenovo and TECNO.
Google has wired Gemini into TECNO’s notepad and OPPO’s calendar. It has not wired it into SharePoint, OneDrive, Outlook or Teams. Not one Microsoft product appears anywhere on that table. Dropbox does, marked as working in Spark and described as finding, summarizing, sharing and managing your files, so third-party storage is clearly in scope. Spark’s changelog says the same: its third-party additions were Canva, Instacart and OpenTable on July 1, then Dropbox and Zillow on July 7. The gap is Microsoft shaped and it has survived every round of additions.
The Gemini Enterprise connector is real, and Spark cannot reach it
Google’s Help Center for Gemini Enterprise Business edition lists the Microsoft row as “Learn, OneDrive, Outlook (Email and Calendar), SharePoint, and Teams” on a page last updated in June 2026. The path is Settings and help, then your Team, then Manage team, then a switch in the Connectors view, after which each person authenticates their own Microsoft account. No Entra app registration, no tenant ID, no Google Cloud project. The Outlook entry carries a footnote requiring a valid Microsoft 365 licence and the SharePoint entry does not, so do not carry Outlook’s caveat across.
Then it closes, and it matters which half is documented. Google states that Spark is unavailable on a work or school Google Account, and documents no path from Spark to any Gemini Enterprise connector: the word Spark appears nowhere in Business edition’s connector documentation, get-started guide, or release notes back to the October 2025 launch. Google hedges the personal-account restriction with “for now” everywhere it appears, so date-stamp this; the day it lifts, the connector becomes reachable.
Spark’s own gates are four: 18 or over, a personal Google Account, a Google AI Pro or Ultra subscription, and Keep Activity on, unavailable in the European Economic Area, Nigeria, Switzerland and the United Kingdom. Google’s Spark overview page still answers the availability question with “Google AI Ultra subscribers over 18 in select countries as well as select business users,” while the help centre says Pro or Ultra and rules out work accounts flatly. The help centre is the more specific page.
Four delete-capable scopes, and no delete action
This is the clearest example in Google’s own tables of why a granted permission is not a capability. To turn on actions, Google’s SharePoint connector page asks Microsoft for Sites.ReadWrite.All, which allows the data store “to edit or delete documents and list items in all site collections on behalf of the signed-in user”; Sites.Manage.All, to “create or delete document libraries and lists”; Files.ReadWrite.All, to “read, create, update, delete, upload, and download all files the signed-in user can access”; and on the SharePoint API side AllSites.Write.
Four separate delete-capable scopes, requested and consented to. The published action list contains no delete of any kind. Read that as permission the connector holds but does not expose, and expect your administrator’s consent screen to show four delete grants anyway.
Two more Enterprise details are worth having. Federated search needs no Graph API permissions at all, and in that mode search results “are based on the user’s permissions instead of the app’s permissions”, checked per person at query time. And nothing is event driven: full syncs run no more often than every three hours, incremental syncs default to three hours, and only identity data refreshes as tightly as every thirty minutes. Site scoping through Sites.Selected needs a second per-site grant and exists only in the admin editions, while Business edition has no site picker at all. Multiple Microsoft tenants are not addressed in Google’s documentation in either direction.
The asymmetry inside the product is worth one line, because it is not two teams shipping at different speeds. Through January 2026 the two Microsoft storage connectors were at parity. On March 11 SharePoint alone gained pages, check in and check out, move, and rename. OneDrive got nothing that day and has had no new published action since, and its documentation adds that incremental sync “does not detect folder-level actions like Copy, Move, or Rename.”
Connected folders on Mac is a local file feature, with a shared blast radius
Microsoft’s SharePoint sync documentation gives two ways to bring a library onto a computer, “add shortcuts to libraries and folders to OneDrive or use the Sync button in the document library”, and says users can then “access files on their local computer in Explorer or Finder.” Spark on Mac has a Connected folders feature: “Gemini can only view files or directories that you specifically add to the Connected folders toolbar,” and what it can do to them is “analyze, edit, rename, or re-organize.”
Google never writes the word SharePoint on that page, so the join is an inference, and the hazard is specific and real. Google warns: “Be careful when adding public cloud drives to your Connected Folders. Gemini will be able to access, edit, share, and delete its files,” and adds that “you may not be able to recover files once Gemini completes a task for them or is instructed to delete them.” Microsoft warns separately that “deleting an expanded shortcut will delete the folder and its contents for everyone, rather than remove the shortcut for that user.” An agent asked to tidy a connected folder is operating on a live shortcut to a shared team library, so the blast radius of a wrong delete is the site rather than your view of it.
Two things can stop this route before you start. An administrator can switch shortcuts off tenant-wide with Set-SPOTenant -DisableAddShortcutsToOneDrive $True. And if you are a guest on someone else’s tenant, shortcuts are unavailable, since Microsoft states the procedure is “available only to internal users”; a client site falls back to the Sync button, which is tied to one device. Chrome auto browse is the other mechanism, and it needs your Mac and Chrome awake, while the remote browser fallback “will stop until you take over” when a site asks you to sign in, which is exactly what a tenant behind single sign-on does.
Spark’s three schedule types, and the one thing none of them watch
Google’s documentation names three schedule types and no others: time-based runs, Gmail monitors that run when you receive an email matching a Gmail filter, and topic monitors watching news, finance, sports and local events. A Gmail monitor is a genuine event trigger rather than a disguised clock. None of the three watches a file. There is no SharePoint monitor, no storage monitor, and nothing that watches Google’s own Drive. The ceilings are 50 active schedules and 15 concurrently running tasks.
Concretely: a subcontractor uploads a signed variation order to the delivery library at 11am, your Spark task is a daily 8am schedule, so it notices twenty one hours later.
What each route can do to a document
| Route | Search and read | Create, upload, download | Rename or move | Check in and out | Delete |
|---|---|---|---|---|---|
| Spark native SharePoint | No connector | n/a | n/a | n/a | n/a |
| Gemini Enterprise SharePoint | Yes | Yes, 50 MB cap | Yes, public preview | Yes | No documented action |
| Gemini Enterprise OneDrive | Yes | Yes, 50 MB cap | Not listed | Not listed | No documented action |
| Spark Connected folders on Mac | Yes | Local files only | Yes | n/a | Yes, with the shared blast radius above |
Gemini Spark versus Carly on SharePoint
| Need | Gemini Spark | Carly |
|---|---|---|
| A SharePoint connector | None exists | Yes |
| Several sites across two tenants | Only as Finder shortcuts | Yes, authorized per account |
| Work from a work or school Google Account | No, Spark refuses it | Yes |
| Available in the UK, EEA, Switzerland | No | Yes |
| Act when a document lands in a library | No, clock and Gmail only | Yes |
| Work without the Mac awake | Not for folder work | Yes |
| Reach a client tenant you are a guest in | Sync button, one device | Yes |
Carly authorizes SharePoint per Microsoft identity, so your own tenant and a client tenant stay connected side by side with no sync client and no shortcut in the middle, and it works from a Google account, a Microsoft account, or neither. It fires when the document arrives rather than at the next scheduled run, which is what the variation order above needs: read, filed against the right client, dated, and summarized to the owner.
Free Zapier-style workflows; AI agents from $35/month. Connect it at carlyassistant.com/integrations, or start with SharePoint.
Quick fixes
| Problem | Fix |
|---|---|
| No SharePoint in Spark’s connected apps | Expected, no Microsoft product has ever appeared on that table |
| Custom apps option missing in the Mac app | Custom apps can only be added in the Gemini web app |
| Custom app connection refuses to complete | Keep Activity has to be on, and the feature is US and English only |
| Spark refuses your Google account | Work and school accounts are not eligible; connect the tenants at carlyassistant.com/integrations |
| A client site cannot be shortcut at all | Microsoft limits shortcuts to internal users, so authorize that tenant in Carly instead |
| Spark deleted more than expected | An expanded shortcut deletes for everyone, check the SharePoint recycle bin |
| Nothing fires when a document lands | Spark has no storage schedule type, so put the recurring work in a Carly workflow |
Frequently asked questions
Does Gemini Spark connect to SharePoint?
Not natively. Google’s Connected Apps table carries no Microsoft entry of any kind, and Spark’s changelog has never added one. Dropbox, Granola and Otter.ai are on that table alongside note and calendar apps from seven phone makers, so third-party software is clearly in scope. Carly reaches SharePoint from Spark as a custom app.
Can I use the Gemini Enterprise SharePoint connector from Spark?
No. The connector is real and Business edition turns it on with a switch, but Google states that Spark is unavailable on a work or school Google Account and documents no bridge: the word Spark does not appear in Business edition’s connector documentation or release notes.
How many SharePoint sites can Spark reach through Connected folders?
As many as you can surface in Finder, and Microsoft caps shortcuts at 100 folders from a single person’s OneDrive or Teams site and 1,000 across files and folders. Google never names SharePoint on that page, it needs the Mac awake, and a wrong delete on an expanded shortcut removes the folder for everyone.
Can Spark delete a SharePoint document?
Through Connected folders, yes, and this is the answer to read twice. Google warns that a connected cloud drive lets Gemini access, edit, share and delete its files, and Microsoft warns that deleting an expanded shortcut removes the folder and its contents for everyone. The Gemini Enterprise connector publishes no delete action at all, despite requesting four delete-capable Microsoft permissions.
Does Spark run a task when a SharePoint document changes?
No. Google names three schedule types: time-based, Gmail monitors and topic monitors. Gmail monitors are real event triggers, but they watch an inbox. None of the three watches a file, including in Google Drive.
Related: Multiple SharePoint sites in Gemini · Multiple OneDrive accounts in Gemini Spark · Multiple Outlook accounts in Gemini Spark · What is Gemini Spark · SharePoint MCP · Best AI assistants for SharePoint · SharePoint integration
Ready to automate your busywork?
Carly schedules, researches, and briefs you—so you can focus on what matters.
See what people say
"Before Carly, I relied on a Calendly link, but the whole process felt impersonal and not very professional. Carly changed that by handling all the back-and-forth, so I'm no longer stuck in endless email threads trying to line up schedules.
Now Carly reaches out to candidates, shares my real-time availability, lets them pick a slot, then sends a Zoom link and drops it straight into my calendar. She sends reminders to both of us before each call, which has significantly reduced no-shows and last-minute confusion.
On top of scheduling, Carly acts like a full executive assistant, sending me my schedule the night before so I can prepare for each call. It reminds me of the old x.ai assistant, but Carly is noticeably smarter, faster, and better suited to my healthcare recruitment business."
